Research on Control Mode Switching of Vehicle Intelligent Suspension Based on DBN and T–S Fuzzy Method

Abstract

To cover the problem of dangerous state prediction ahead of vehicle rollover and rescue the vehicle under abrupt cornering condition, a dynamic Bayesian network (DBN) merged robust control is developed to balance the vehicle ride comfort and handling performance. To discretize the automobile state attributes and prepare for the prediction, class attribute contingency coefficient (CACC) is adopted to pre-process the data conveniently and establish the category labels. The key contributions of this paper are efficient rollover prediction with probabilistic and numerical representation, a mapping rule from rollover probabilities to T–S fuzzy membership values, and an intelligent objective switchable control between ride comfort and roll stability. The co-simulation method is adopted to verify the effectiveness of this method with passive suspension, semi-active suspension, and optimal control active suspension. It is shown that the DBN-based robust control is able to reduce the roll angle by more than 27% compared to the passive suspension under double-lane change condition and has the best balancing performance. From the perspective of ride comfort testing on bounce sinusoidal roads, the vehicle DBN incorporating robust controllers can effectively reject vibrations and switch control objectives based on its running conditions.